TMW->ACIP is much improved. V and W were confused, # and * were

confused; many glyphs that should have yielded errors were not.

I've added a test case that transforms every TMW glyph save the one with
no TM mapping to ACIP.  I hand-checked that it was correct.

ACIP->TMW is fixed for # and *.  I never noticed it, but each needed an
extra swoosh (U+0F05).

Round-tripping would be good, as would testing real-world use of
TMW->ACIP.

source/org/thdl/tib/text/ttt/TParseTree.java

@@ -359,6 +359,7 @@ class TParseTree {
                         }
                     }
                     if (stackSize > 1 && tp.getLeft() != null && tp.getLeft().length() > 1) {
+                        // DLC FIXME: gives a false positive warning for Rsh
                         hasAmbiguousConsonant = true;
                     }
                 }